20201019 ML Brief
If you’re not updating ROAS and CPA targets frequently, you’re missing out
Our founding CEO, Fred Vallaeys, has written extensively about how the role of PPC experts will change as more of the day-to-day ad management will be done by automated systems driven by artificial intelligence (AI) and machine learning (ML). One task that’s particularly apt to become automated is bid management because it requires lots of math and pattern recognition, two tasks at which computers excel.
But as much as automated account management is improving thanks to increased computing power, setting the perfect bid sometimes requires something that the computers don’t have: deep insight into your business. We believe PPC pros can boost their results by combining their unique insights into the business with the massive computation power of Google and Bing’s AI algorithms.
The benefits of automated bidding
Google has explained in a white paper how its automated system calculates bids. It analyzes auction-time signals to predict the likely conversion rate and conversion value from a click. It does this using a massive amount of data from all ads, advertisers and the billions of clicks they get. Individual advertisers don’t have nearly as much data, let alone the power of supercomputers to analyze and act on it in real time. So leveraging Google’s AI is a smart thing to do for the typical search marketer who’s looking to improve their performance.
Image courtesy of Google.com. Factors analyzed for automated bidding.
But as great as Google’s bidding system is, the factors it analyzes are limited. So, while it might be able to predict the difference in the likelihood of a conversion from a user located in Germany at 2 o’clock in the afternoon versus from a user located in Canada at 7 o’clock in the morning, it doesn’t know about other unique aspects of the business that impact what results are likely to come from a click.
There are potentially thousands of factors that the algorithm doesn’t consider, simply because the data for those factors might be too sparse, too inconsistent between different types of advertisers or too expensive to include in the prediction models.
The benefits of human-assisted bids
Humans who manage PPC accounts can use their intuition to make some pretty good guesses about how these unique factors might impact results.
For example, a boat shop could reasonably expect to rent more Jet Skis when the following are true:
- The occupancy rate of hotels near the beach is high.
- The forecast calls for lots of sun and calm winds.
- There hasn’t been a news story about shark attacks.
Google’s automated bidding doesn’t consider these factors, so it means that its bids, while good, won’t be perfect. Human account managers, on the other hand, can change bids based on their experience of how these factors normally affect rentals.
The problem is that you don’t really know if the factors Google considers are more important to performance than the ones humans would consider. But there is a way to measure the difference.
How to test if automated bids are better
To test if automated bidding delivers better performance than manual CPC bidding, advertisers can use Google’s Drafts and Experiments.
Take an existing campaign that’s on manual bidding or where automation is handled by a third-party tool, and create a new draft campaign from it. In the draft, change only the bidding strategy, and then launch it as an experiment.
Image courtesy of Google.com.
How much traffic the campaign gets and how much of that traffic is sent to the experiment will determine how soon results will be available. Once you find the winning bid strategy, you can end the experiment and revert to the original bidding method or adopt the new one.
The best bids are automated, but with lots of human input
It’s possible to get even better results by combining the strengths of humans and machines. After all, nobody said that automated bids are a set-it-and-forget-it method.
Advertisers tend to think of automated bidding as the end of their work on bid management, but we believe that target CPA and target ROAS should be actively managed. Instead of using the max CPC as a lever for optimization, target ROAS and target CPA can play that same role.
When managing CPC bids, advertisers change bids based on factors like changes in ad performance or how factors outside AdWords impact the lead-to-sales conversion rate. Likewise, ROAS and CPA targets should be managed actively based on what is happening in AdWords and with the business.
Google recommends changing CPA and ROAS targets
Even Google says this: “To accommodate … brief, anticipated changes in performance, we recommend that advertisers adjust their bidding targets (target CPA or target ROAS) proportionately to the predicted increase or decrease in conversion rate or value.”
Google is saying the algorithms need time to update their predictions and for best performance, humans should tweak the targets during periods of flux when Google’s expected conversion rate will not be close to the actual conversion rate. For example, if a special sale is expected to dramatically increase the conversion rates, advertisers should change their targets if they don’t want to miss out on an opportunity to sell more.
Here’s a simplified explanation of how this plays out with automated bids.
If a sale starts on Day 4, the actual conversion rate may instantly double, but the predicted conversion rate could take some time to catch up with the new reality.
During the time when the new actual conversion rate is very different from the old predicted conversion rate, it is likely the automated bids will still reflect the old conversion rate, and hence be incorrect (this assumes the advertiser does not change their CPA target).
Automate managing target ROAS and target CPA with Optmyzr
Fortunately, getting the best of Google’s automated bids with a layer of human intelligence doesn’t mean you have to go back to managing things manually. Tools like Optmyzr’s Rule Engine can automatically adjust target ROAS and target CPA values based on your own calculations.
A Rule Engine can be used to adjust target ROAS and target CPA bids automatically based on conditions specified by advertisers, even using data from outside AdWords.
The Optmyzr system can even be connected with your own business data, whether from a CRM, a database of product margins, a promotional calendar, the weather or something else. This data that is not available inside AdWords can then be used to make better decisions about the targets the automated Google system should hit.
Conclusion
While automated bidding tools from the search engines can take away the absolute need for human account managers to spend time on this task, there are good reasons why smart PPC pros will still want to be actively involved in changing their targets. Whereas in the past, bid management was all about changing max CPC bids, it’s now about changing target ROAS and target CPAs to optimize accounts. Tools like Optmyzr make it easier to benefit from Google’s AI while still maintaining that human touch in account management.
The post If you’re not updating ROAS and CPA targets frequently, you’re missing out appeared first on Marketing Land.
From our sponsors: If you’re not updating ROAS and CPA targets frequently, you’re missing out
